SparsePipe: Parallel Deep Learning for 3D Point Clouds

TL;DR

SparsePipe introduces an efficient parallel training framework for 3D point cloud neural networks, leveraging sparse tensor representations and optimized GPU parallelism to improve performance and reduce memory usage.

Contribution

It presents a novel asynchronous parallelism approach tailored for 3D sparse data, enabling high-resolution processing and efficient multi-GPU training.

Findings

Achieves higher training throughput on multi-GPU setups.

Reduces memory requirements compared to dense models.

Improves performance on point cloud benchmarks.

Abstract

We propose SparsePipe, an efficient and asynchronous parallelism approach for handling 3D point clouds with multi-GPU training. SparsePipe is built to support 3D sparse data such as point clouds. It achieves this by adopting generalized convolutions with sparse tensor representation to build expressive high-dimensional convolutional neural networks. Compared to dense solutions, the new models can efficiently process irregular point clouds without densely sliding over the entire space, significantly reducing the memory requirements and allowing higher resolutions of the underlying 3D volumes for better performance. SparsePipe exploits intra-batch parallelism that partitions input data into multiple processors and further improves the training throughput with inter-batch pipelining to overlap communication and computing. Besides, it suitably partitions the model when the GPUs are heterogeneous such that the computing is load-balanced with reduced communication overhead. Using experimental results on an eight-GPU platform, we show that SparsePipe can parallelize effectively and obtain better performance on current point cloud benchmarks for both training and inference, compared to its dense solutions.